TL;DR

• Enterprise applications are bundles of data model, business logic, UI, and workflow. AI agents only need the first two. The UI and workflow layers are losing strategic importance, and the enterprise stack is reorganizing around a new three-layer architecture (Enterprise Platform, Agentic Platform, Collaboration) with cross-cutting vertical services for Identity, Governance, Context/Memory, and Metering.

• The Agent Service Bus, the connective tissue between agent platforms and collaboration layers, is the most strategically important infrastructure nobody is building. It must perform five functions (capability discovery, intent resolution, contract negotiation, conflict arbitration, message routing), and current protocols (A2A, MCP) address only two.

• The native-versus-external agent debate is a false binary. Native agents offer depth within a vendor's ecosystem; external agents offer breadth across systems. The market is converging on a third path: native agents with open interoperability through A2A and MCP.

• Agentic identity and governance are architectural layers, not security features. Identity requires four dimensions (authentication, authorization, accountability, provenance) that human-centric IAM cannot address. Governance requires five layers (access, compliance, behavioral, contextual, accountability) that go well beyond what most vendors offer today.

• Per-seat pricing is structurally broken. Value-based pricing fails in practice (attribution, measurement, adversarial dynamics, unpredictability). Hybrid models, a platform fee plus consumption metering, are the practical path forward. Metering infrastructure is the prerequisite most enterprises lack.

• Cross-organizational agent collaboration is where the entire architecture gets stress-tested. Identity must be federated, governance must be bilateral, orchestration must bridge separate infrastructures. The likely outcome is a hybrid of vertical platforms for industry depth and horizontal protocols for cross-industry breadth.

• The center of gravity shifts over time in three overlapping phases: data wins near-term (favoring Oracle, SAP, Salesforce), intelligence wins mid-term (favoring OpenAI, Anthropic, Microsoft), and business logic wins long-term (favoring whoever owns it in machine-readable, platform-independent form). The timeline is accelerating.

• The single most important long-term strategic action for enterprises: document, structure, and own your business logic independently of vendor platforms. The enterprises that do this will have options. The enterprises that leave their business logic buried inside vendor applications will not.